Can you expand on this for me. I can see that for any input to a dac something is going to have to happen when each bit is received- presumably this would be equally true for S/PDIF or usb -something has to happen when a bit is read(?)
SPDIF is a far simpler interface with significantly simpler decoding circuits, and can be decoded at a fixed Clock rate with oversampled statistical analysis (IIRC so long as you can sample the edges at 3 - 4 times the data rate you can recover the Data "clockless".
The USB interface requires magnitudes greater circuit complexity, the more circuit devices you have operating at the USB Host data rate the more greater the level of PSU & Ground noise will can be injected into the system. This opens the door for a "Distortion" mechanism that allows the PC's operating domain to enter the Audio domain.
The Ideal solution is to remotely situate the USB input circuit external to the DAC chassis, and connect via optical links. The DAC can provide the Reference Master clock so the USB port is operated Asynchronous in the correct sense.
Is this problem specific to memory buffers or does it remain with any digital transmission? Is there any particular reason why the little spark of energy that occurs when the incoming bit is read into the buffer will cause more problems than when an incoming bit is received by the usb receiver, or I suppose when optical data is converted to electrical signal?
Only in the fact that greater circuit complexity,the greater radiated energy levels, clocking data into a deep memory buffer means many cells operating at the income data rate, USB requires Memory buffer circuits and communications at the Hosts data rate.... all circuits that are clocked at the Jittery input clock rates.
Is the problem here that if the incoming signal is more jittery there is a greater chance that the second order effect will be damaging. -I wasn't quite clear whether the transmission of this little glitch of energy might be equally problematic whether it was jittery or not. This problem intrigues me because it makes me wonder whether usb might be worse than S/PDIF: would it be transmitted evenly or in more jittery packets than S/PDIF.
Yes exactly this - the greater the input jitter, the greater the noise spectrum components produced, these components will leak into the analogue domain within such a non screen enclosure. We can only attenuate these levels, but never attenuate to zero.
Is it possible that you could alleviate the problem by somehow dithering the input of the data to de-correlate whatever second order effect there was?
Yes, very much so, and the reason the MDAC2 has the advanced clock option where we can eject a controlled amount and spectrum of "Jitter" to de-correlate such effects.
A problem with SPDIF and to a degree USB is a that the "Energy spectrum" of the Data is strongly correlated to the transmitted Audio data. If I where to design the SPDIF interface today (Silicon area is so much cheaper today then when SPDIF was first conceived) I'd scramble the Audio data during transmission to de-correlate such second order effects. This would also greatly help the Clock recovery circuits (PLL's) who have a hard time attenuating these correlated audio data patterns.